Skip to content

Holotorch is an optimization framework for differentiable wave-propagation written in PyTorch

License

Notifications You must be signed in to change notification settings

facebookresearch/holotorch

Repository files navigation

Holotorch

Holotorch is a Fourier Optics / Coherent Imaging framework developped in PyTorch and PyTorch lightning.

The main functions are found in the folders "holotorch".

Holotorch provides the following:

  1. Simple setup of optical setups in simulation for forward modeling
    1. Optical propagators (e.g. ASM-Kernel, Optical fouriertransforms)
    2. Abstracted componenents like Source, Detector, SLMs, DoEs, etc.
  2. Complex Wavefront objects which carry more information than just a "datatensor", such as spectral and spatial information
  3. Automatic batching and saving/loading of SLM-states for more complex "Machine Learning" tasks for "Deep Optics"
  4. Abstracted code for hardware often used in research holographic displays
    1. Cameras (Flir + Ximea)
    2. SLM ( when SLM is displayed as second screen. Based on python package slmpy )
  5. Data aquisition pipelines to capture datasets that are e.g. useful for calibration procedures (e.g. Neural Holography)
  6. PyTorch Lightning Modules tailored for Computational Holography
    1. SLM-Lightning: Simple optimization algorithm based on gradient descent where we optimize for an SLM given an optical setup (e.g. Near-Eye or Far-Field)
    2. (Neural) Etendue Lightning: Joint optimization of a hologram and SLM-patterns

Best Practice

see holotorch_and_visual_studio_code.md

Example Notebook

Please navigate to SIGGRAPH_Tutorial and open "tutorial.ipynb"

The Siggraph tutorial notebook contains:

  1. Example for ASM-propagator
  2. Creating a Hologram using Double-Phase-Amplitude-Encoding (DPAC)
  3. Near-Eye Hologram (ASM-propagation) optimization
  4. Conventional Etendue Expansion with a random diffuser
  5. Neural Etendue expansion (Deep Optics) with respect to an image dataset
  6. Examples on how to save/load learned models

Contact

Contact: florianschiffers ( at ) gmail.com or florian.schiffers ( at ) u.northwestern.edu ocossairt ( at ) fb.com nathanmatsuda ( at ) fb.com

LICENSE

Shield: CC BY-NC-SA 4.0

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

CC BY-NC-SA 4.0

About

Holotorch is an optimization framework for differentiable wave-propagation written in PyTorch

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published